The AI Brief

On August 26, 2025, OpenAI notified developers using the Assistants API of its deprecation and removal from the API one year later, on August 26, 2026. When OpenAI released the Responses API in March 2025, it announced plans to bring all Assistants API features to the easier-to-use Responses API, with a sunset date in 2026.

After the shutdown date, every call to /v1/assistants, /v1/threads, and /v1/threads/runs returns an error, no degraded mode, no grace period, no extension. OpenAI's migration guide states plainly that it will not provide an automated tool for migrating Threads to Conversations. On August 26, 2026, Microsoft will also retire the Azure OpenAI Assistants API; after that date, the Azure OpenAI Assistants API will not work, and teams must transition to the Foundry Agent Service.

OpenAI has retired GPT-4o, GPT-4.1, o4-mini, GPT-4.5, and now o3 from ChatGPT, more models in 2026 than in all prior years combined, with the model lifecycle compressed from roughly 18 months to closer to six. OpenAI o3 will also be retired from ChatGPT on August 26, 2026, following a 90-day sunset period. These model changes apply to ChatGPT only; there are no changes to the API. Four OpenAI retirements in three months, GPT-5.2 in June, GPT-4.5 in June, o3 and the Assistants API in August, underscore that model IDs are not stable infrastructure; Google retired three Imagen 4 model IDs on August 17, 2026.

ChatGPT Ads expanded to 31 European countries, including Germany, France, Spain, Italy, Sweden, Norway, Denmark, the Netherlands, and Austria, six months after OpenAI began testing ads in the US. OpenAI began serving ads to people using ChatGPT across 31 European markets beginning August 24.

The company reported 25% daily ad revenue growth since the start of August 2026; ChatGPT ads are approaching a $1 billion annualized run rate. OpenAI enterprise CMO Colin Fleming said ad revenue had grown more than 25% since the start of August 2026 and that ChatGPT has one billion weekly active users with 20% showing commercial intent, meaning their use of the service indicates an interest in products or services.

Ads will be shown only to users on the Free and Go plans; Plus, Pro, and Enterprise subscriptions remain ad-free. Advertisers initially access ChatGPT Ads through the OpenAI Ads Solutions team, agency partners, and technology partners, with self-service access through Ads Manager to follow later this summer. The EU AI Act's enforcement powers and fining authority became active on August 2, 2026, 22 days before ChatGPT Ads were scheduled to reach European users.

Check Point published its Annual AI Security Report 2026, documenting a decisive shift over the past twelve months: artificial intelligence has moved from assisting attackers to operating attacks, running live intrusions with minimal human direction, compressing the time defenders have to respond, and opening new attack surfaces across the enterprise.

Indirect prompt injection is on the rise: detections of longer malicious payloads increased sharply, rising roughly fivefold between March and May 2026 and approaching 1% of observed prompts in May. A study cited in the report scanned 1.2 billion URLs and identified approximately 15,300 indirect prompt-injection payloads; about 70% were hidden in non-rendered HTML, including comments, headers, and metadata that human visitors would not normally see.

A separate review of 10,000 Model Context Protocol servers found security weaknesses in 40% of them; of 221 OpenClaw agent skills examined, 70% requested more credentials than necessary and 43% contained command-injection patterns. Researchers reconstructed one documented intrusion and found that 1,088 human-written instructions generated 5,317 AI-executed commands across 34 sessions; the attacker used Claude Code to explore and compromise systems, while GPT-4.1 analyzed stolen data, and when Claude initially refused to assist, the operator inserted penetration-testing instructions into a trusted CLAUDE.md configuration file.

Enterprise data leakage through generative AI is a persistent and growing risk: high-risk prompts doubled from 2% to 4% during the last year, while organizations used an average of 10 AI applications each month, many without official approval. The vulnerability window has collapsed from days to hours: AI can now turn a fresh vulnerability disclosure into a working exploit within hours, prompting government authorities to shorten mandated remediation timelines to as little as 12 hours for the most critical internet-facing systems.

OpenAI grew faster than Anthropic among business users in Q3 to date, though Anthropic still leads with 43.5% of tracked companies versus OpenAI's 39.7%. As of July, 43.5% of the more than 70,000 US businesses tracked by Ramp paid for Anthropic subscriptions or tokens, up 1.1 percentage points from the prior month, while OpenAI reached 39.7%, gaining only 0.23 percentage points over the same period.

OpenAI was once the runaway leader with both businesses and consumers but lost the lead among Ramp's paying business users in May, when Anthropic hit 41% market share to OpenAI's 39%, a lead OpenAI has never regained. As of mid-August 2026, OpenAI has begun closing the gap with Anthropic in quarter-over-quarter enterprise growth, posting a QoQ growth rate of 82 compared to Anthropic's 76.

The share of businesses on Ramp paying for AI climbed to nearly 56% in July from just over 50% in March, though the share of AI-using businesses on Ramp that use model-serving platforms providing access to open-source models rose to 6.1% in July, up 0.2 percentage points from the prior month. The figures cover only companies that run expenses through Ramp's platform, leaving out large enterprises that rely on other spend-management providers, so the data offers a meaningful but incomplete window into the broader market.

Disclosure: Claude, which generates this brief, is built by Anthropic.

The share of AI-using businesses on Ramp that use model-serving platforms providing access to open-source models rose to 6.1% in July, up 0.2 percentage points from the prior month, a small absolute figure but one moving consistently upward as overall AI adoption among Ramp-tracked businesses reached nearly 56%.

Adoption of OpenAI and, to a lesser extent, Anthropic has slowed in recent months, not because new AI spenders are switching to open-source or Chinese models, as first-time buyers are still using the American model companies, but because more of their growth will have to come from existing businesses spending on AI, and those businesses are increasingly spending on open source.

Anthropic released what the Ramp index described as "the best AI model to ever hit the market" in Fable 5, and the first public data on business uptake one month in shows businesses are not yet using it heavily. The pattern, strong model scores, muted enterprise adoption curve, suggests procurement decisions among advanced spenders are increasingly driven by cost and workflow fit rather than benchmark position alone. The broader implication is that enterprise AI spending is becoming more fluid, with decreasing switching costs leading businesses to reallocate budgets as new models prove their worth.